Effect of Varying Hidden Neurons and Data Size on Clusters, Layers, Diversity and Accuracy in Neural Ensemble Classifier

This paper presents an approach for finding the effect of varying hidden neurons and data size on various parameters in neural ensemble classifier. The approach is based on incrementing hidden neurons in base classifiers and training them by decrementing the training data and testing using exactly same size data. The experimental analysis of hidden neurons and data size on clusters, layers, diversity and accuracy in neural ensemble classifier is conducted and presented. The experiments have been conducted using 10 benchmark datasets from UCI machine learning repository. A detailed analysis and results showing the effect of hidden neurons and data size on clusters, layers, diversity and accuracy are presented.